LOOMAL
AI NPS Survey Agent

NPS that actually
drives action.

Most NPS programs are dead before they start. The survey goes out, scores come back, a dashboard updates, and nothing changes. A Loomal agent runs the entire loop — sending, reading, classifying, and reaching back out — so feedback turns into save plays and advocacy moments instead of slide-deck filler.

mail_sendmail_list_messagesmail_reply

API Primitives used

mail_send

Send the survey

Outbound from a CS agent identity, DKIM-signed, threaded against the customer's main contact — not a third-party survey domain.

mail_list_messages

Catch responses

Replies land in the agent inbox. extractedText gives the LLM a clean view of the score and the comment underneath it.

mail_reply

Follow up in-thread

Detractors get a fast, human-feeling apology and escalation. Promoters get an advocacy ask. All inside the original conversation.

NPS dies in the dashboard.

Every CS team runs NPS. Almost none of them act on it. The survey tool ships responses to a dashboard, the dashboard ships a number to a slide, and the slide goes into a quarterly review. Meanwhile the detractor who took the time to write a comment gets nothing back — confirming exactly the experience that scored them low in the first place.

The fix isn't a better dashboard. It's a real human-feeling response loop that runs at every score, every time, without burning a CSM's afternoon. That's an agent job, and it requires an inbox, a model, and an audit trail in one identity.

How to build it.

01mail_send

Send the survey

Agent sends the NPS request from the CS identity at the right moment — post-onboarding, post-renewal, or post-support resolution.

02mail_list_messages

Read the reply

Agent parses the score and the comment, classifies sentiment, and decides whether to escalate, advocate, or thank-and-close.

03mail_reply

Respond in-thread

Detractors get a real apology and a CSM call booked. Passives get a thoughtful follow-up question. Promoters get an advocacy ask.

Example prompt

Send NPS to every customer 30 days post-onboarding. Reply to any score under 7 with an apology and book a call with their CSM. Reply to scores of 9 or 10 asking if they'd write a review.

What CS teams build.

Lifecycle-triggered surveys

Agent sends NPS at key moments — post-onboarding, post-resolution, post-renewal — instead of quarterly blasts.

Detractor save plays

Low scores trigger an immediate apology, a real call invite, and an internal flag for the CSM with the customer's exact words attached.

Promoter advocacy

High scores prompt review requests, case-study asks, and referrals — caught in the moment of peak goodwill.

Comment classification

Agent reads open comments and tags them by theme — pricing, support, feature gap — feeding the product team real signal.

Multi-segment surveys

Different customer tiers get different surveys from different agent identities, with full audit per segment.

Why NPS belongs to an agent.

NPS is a perfect example of a workflow that's high-value, high-volume, and chronically under-resourced. The action items hide in long-tail comments that nobody has time to read. An agent reads all of them, classifies them, and acts — at a quality level humans would aspire to but can't sustain.

Loomal's identity model means every survey, every reply, and every escalation is tied to a real, audited agent — not a third-party survey tool that owns the data and the relationship. The customer never leaves your domain; the feedback never leaves your trust boundary.

Customer data stays internal

Survey responses live in your own inbox, not a third-party survey vendor.

Per-survey audit

Every send, classification, and reply is logged with the agent identity and originating event.

Identity-scoped sends

Each customer segment can be surveyed by a different agent identity for clean reporting and revocation.

Related use cases

Make NPS feedback drive action.

Send, classify, and respond — fully autonomous.